Machine Learning with Multiple Constraints

1. A computer-implemented method comprising: detecting a user uploading data and one or more constraints; collecting the data and the one or more constraints; and determining that one or more model pipelines satisfies all of the one or more constraints based on applying one or more algorithms to the collected data, the collected constraints, and to multiple model pipelines, wherein the one or more algorithms use an alternating direction method of multipliers (ADMM) based joint optimization approach in determining that one or more of the model pipelines satisfies all of the one or more constraints, and the ADMM based joint optimization approach decomposes the determining into sub-problems solved iteratively, the sub-problems comprising: one or more hyperparameter problems solving hyperparameter optimization for the one or more model pipelines, and a pipeline selection problem solved by a combinatorial multi-armed bandit algorithm,, wherein the ADMM based joint optimization approach uses explicit Boolean variables to encode selection of the multiple model pipelines.

2. The method of claim 1, further comprising: applying the determined one or more model pipelines to the collected data.

3. The method of claim 2, further comprising: notifying the user of the determined one or more model pipelines before applying the determined one or more model pipelines to the collected data.

4. The method of claim 1, wherein: the one or more algorithms re-optimize the one or more hyperparameters based on the selected model pipelines.

5. The method of claim 1, further comprising: receiving feedback indicative of whether the determined one or more model pipelines appropriately satisfied all of the constraints; and adjusting the one or more algorithms based on the received feedback.

6. The method of claim 1, wherein: the one or more constraints include constraints selected from a group consisting of a maximum inference time, maximum false positive rate, maximum false negative rate, maximum memory overhead, maximum parity difference, maximum error rate difference, and minimum area under a receiver operating characteristic curve; and the one or more model pipelines include model pipelines selected from a group consisting of eXtreme Gradient Boosting classifier models, extra trees classifier models, quadratic discriminant models, Gaussian Naive Bayes models, and K nearest neighbors classifier models.

7. The method of claim 1, wherein the determining determines that more than one of the model pipelines satisfies all of the one or more constraints; and wherein the method further comprises applying each of the determined more than one model pipelines to the collected data in parallel.

8. The method of claim 1, wherein the determining comprises the one or more algorithms decomposing an analysis to three smaller problems that are solved iteratively.

9. The method of claim 1, wherein the ADMM based joint optimization approach implements variables to adaptively penalize constraint violations at different stages of the one or more algorithms.

10. A computer program product comprising: one or more non-transitory computer-readable storage media and program instructions stored on the one or more non-transitory computer-readable storage media capable of performing a method, the method comprising: detecting a user uploading data and one or more constraints; collecting the data and the one or more constraints; and determining that one or more model pipelines satisfies all of the one or more constraints based on applying one or more algorithms to the collected data, the collected constraints, and multiple model pipelines, wherein the one or more algorithms use an alternating direction method of multipliers (ADMM) based joint optimization approach in determining that one or more of the model pipelines satisfies all of the one or more constraints, and the ADMM based joint optimization approach differentiates continuous and discrete hyperparameter configurations defined for the multiple model pipelines.

11. The computer program product of claim 10, further comprising: applying the determined one or more model pipelines to the collected data.

12. The computer program product of claim 11, further comprising: notifying the user of the determined one or more model pipelines before applying the determined one or more model pipelines to the collected data.

13. The computer program product of claim 10, wherein: the one or more algorithms optimize one or more hyperparameters and select model pipelines based on the one or more hyperparameters; and the one or more algorithms re-optimize the one or more hyperparameters based on the selected model pipelines.

14. The computer program product of claim 10, further comprising: receiving feedback indicative of whether the determined one or more model pipelines appropriately satisfied all of the constraints; and adjusting the one or more algorithms based on the received feedback.

15. The computer program product of claim 10, wherein: the one or more constraints include constraints selected from a group consisting of a maximum inference time, maximum false positive rate, maximum false negative rate, maximum memory overhead, maximum parity difference, maximum error rate difference, and minimum area under a receiver operating characteristic curve; and the one or more model pipelines include model pipelines selected from a group consisting of eXtreme Gradient Boosting classifier models, extra trees classifier models, quadratic discriminant models, Gaussian Naive Bayes models, and K nearest neighbors classifier models.

16. A computer system for determining that one or more model pipelines satisfy one or more constraints, the computer system comprising: one or more computer processors, one or more computer-readable storage media, and program instructions stored on the one or more of the computer-readable storage media for execution by at least one of the one or more processors to cause a method to be performed comprising: detecting a user uploading data and one or more constraints; collecting the data and the one or more constraints; and determining that one or more model pipelines satisfies all of the one or more constraints based on applying one or more algorithms to the collected data, the collected constraints, and to multiple model pipelines, wherein the one or more algorithms use an alternating direction method of multipliers (ADMM) based joint optimization approach in determining that one or more of the model pipelines satisfies all of the one or more constraints, and the ADMM based joint optimization approach decomposes the determining into sub-problems solved iteratively, the sub-problems comprising: one or more hyperparameter problems solving hyperparameter optimization for the one or more model pipelines, and a pipeline selection problem solved by a combinatorial multi-armed bandit algorithm,, wherein the ADMM based joint optimization approach uses explicit Boolean variables to encode selection of the multiple model pipelines.

17. The computer system of claim 16, further comprising: applying the determined one or more model pipelines to the collected data.

18. The computer system of claim 17, further comprising: notifying the user of the determined one or more model pipelines before applying the determined one or more model pipelines to the collected data.

19. The computer system of claim 16, wherein: the one or more algorithms re-optimize the one or more hyperparameters based on the selected model pipelines.

20. The computer system of claim 16, further comprising: receiving feedback indicative of whether the determined one or more model pipelines appropriately satisfied all of the constraints; and adjusting the one or more algorithms based on the received feedback.